liger-kernel-nightly 0.4.1.dev20241113011623__tar.gz → 0.4.1.dev20241114155849__tar.gz

{liger_kernel_nightly-0.4.1.dev20241113011623/src/liger_kernel_nightly.egg-info → liger_kernel_nightly-0.4.1.dev20241114155849}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: liger_kernel_nightly
-Version: 0.4.1.dev20241113011623
+Version: 0.4.1.dev20241114155849
 Summary: Efficient Triton kernels for LLM Training
 License: BSD 2-CLAUSE LICENSE
         Copyright 2024 LinkedIn Corporation 

{liger_kernel_nightly-0.4.1.dev20241113011623 → liger_kernel_nightly-0.4.1.dev20241114155849}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "liger_kernel_nightly"
-version = "0.4.1.dev20241113011623"
+version = "0.4.1.dev20241114155849"
 description = "Efficient Triton kernels for LLM Training"
 urls = { "Homepage" = "https://github.com/linkedin/Liger-Kernel" }
 readme = { file = "README.md", content-type = "text/markdown" }

liger_kernel_nightly-0.4.1.dev20241114155849/src/liger_kernel/chunked_loss/fused_linear_preference.py

@@ -0,0 +1,206 @@
+from abc import abstractmethod
+from functools import partial
+
+import torch
+from torch.nn import functional as F
+
+
+class LigerFusedLinearPreferenceBase(torch.autograd.Function):
+
+    @abstractmethod
+    def preference_loss_fn(chosen_logps, rejected_logps, beta=0.1):
+        """
+        Compute preference loss.
+        Args:
+            chosen_logps (torch.Tensor): Avg log probabilities of chosen tokens. Shape: (batch_size,).
+            rejected_logps (torch.Tensor): Avg log probabilities of rejected tokens. Shape: (batch_size,).
+            beta (float): Weight for the odds ratio loss.
+        """
+        raise NotImplementedError("Preference loss function must be implemented.")
+
+    @staticmethod
+    def forward(
+        ctx,
+        _input,
+        weight,
+        target,
+        bias=None,
+        loss_fn=None,
+        chunk_size=1,
+        compute_nll_loss=True,
+        ignore_index=-100,
+        beta=0.1,
+        compiled=True,
+    ):
+        """
+        Base class for fused linear layer with preference loss.
+        Expects _input to be stacked with chosen and rejected inputs on the batch dimension.
+
+        Args:
+            _input (torch.Tensor): Input tensor. Shape: (batch_size, seq_len, hidden_size).
+            weight (torch.Tensor): Weight tensor. Shape: (vocab_size, hidden_size).
+            target (torch.Tensor): Target tensor. Shape: (batch_size, seq_len).
+            bias (torch.Tensor, optional): Bias tensor. Shape: (vocab_size,).
+            loss_fn (callable): Loss function to compute the loss on a chunk of input/target.
+            chunk_size (int): Size of a chunk (# of batches of stacked chosen and rejected inputs).
+            compute_nll_loss (bool): Whether to compute NLL loss.
+            ignore_index (int): Index to ignore for loss computation.
+            beta (float): Weight for the odds ratio loss.
+            compiled (bool): Whether to use torch compile for chunk accumulation.
+        """
+        # TODO: Tune CHUNK_SIZE to fully utilize the GPU
+        CHUNK_SIZE = chunk_size
+
+        grad_weight = torch.zeros_like(weight)
+        grad_chosen_inputs = []
+        grad_rejected_inputs = []
+        grad_bias = torch.zeros_like(bias) if bias is not None else None
+        loss_acc = torch.zeros((), device=_input.device)
+
+        chunks = max(1, _input.shape[0] // (2 * CHUNK_SIZE))
+        loss_func_to_call = partial(
+            LigerFusedLinearPreferenceBase._compute_loss,
+            preference_loss_fn=loss_fn,
+            ignore_index=ignore_index,
+            beta=beta,
+            compute_nll_loss=compute_nll_loss,
+            full_target=target,
+        )
+
+        def accumulate_chunk(input_chunk, target_chunk):
+            if bias is not None:
+                (chunk_grad_input, chunk_grad_weight, chunk_grad_bias), (
+                    chunk_loss,
+                    (chunk_or_loss, chunk_chosen_logps, chunk_rejected_logps),
+                ) = torch.func.grad_and_value(
+                    loss_func_to_call, argnums=(0, 1, 3), has_aux=True
+                )(
+                    input_chunk, weight, target_chunk, bias
+                )
+                grad_bias.add_(chunk_grad_bias)
+            else:
+                (chunk_grad_input, chunk_grad_weight), (
+                    chunk_loss,
+                    (chunk_or_loss, chunk_chosen_logps, chunk_rejected_logps),
+                ) = torch.func.grad_and_value(
+                    loss_func_to_call, argnums=(0, 1), has_aux=True
+                )(
+                    input_chunk, weight, target_chunk
+                )
+            grad_weight.add_(chunk_grad_weight)
+            loss_acc.add_(chunk_loss)
+            return chunk_grad_input
+
+        len_chosen = target.shape[0] // 2
+        _chosen_input_chunks = torch.chunk(_input[:len_chosen], chunks=chunks, dim=0)
+        _chosen_target_chunks = torch.chunk(target[:len_chosen], chunks=chunks, dim=0)
+        _rejected_input_chunks = torch.chunk(_input[len_chosen:], chunks=chunks, dim=0)
+        _rejected_target_chunks = torch.chunk(target[len_chosen:], chunks=chunks, dim=0)
+
+        for (
+            chosen_input_chunk,
+            rejected_input_chunk,
+            chosen_target_chunk,
+            rejected_target_chunk,
+        ) in zip(
+            _chosen_input_chunks,
+            _rejected_input_chunks,
+            _chosen_target_chunks,
+            _rejected_target_chunks,
+        ):
+            input_chunk = torch.cat([chosen_input_chunk, rejected_input_chunk], dim=0)
+            target_chunk = torch.cat(
+                [chosen_target_chunk, rejected_target_chunk], dim=0
+            )
+
+            if compiled:
+                accumulate_chunk = torch.compile(accumulate_chunk)
+            grad_input = accumulate_chunk(input_chunk, target_chunk)
+
+            grad_chosen_inputs.append(grad_input[: chosen_target_chunk.shape[0]])
+            grad_rejected_inputs.append(grad_input[chosen_target_chunk.shape[0] :])
+
+        # combine grad_chosen_inputs and grad_rejected_inputs
+        grad_inputs = grad_chosen_inputs + grad_rejected_inputs
+
+        ctx.save_for_backward(
+            torch.cat(grad_inputs, dim=0),
+            grad_weight,
+            grad_bias,
+        )
+        return loss_acc
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        grad_input, grad_weight, grad_bias = ctx.saved_tensors
+        if torch.ne(grad_output, torch.tensor(1.0, device=grad_output.device)):
+            grad_input = grad_input * grad_output
+            grad_weight = grad_weight * grad_output
+            grad_bias = grad_bias * grad_output if grad_bias is not None else None
+
+        return grad_input, grad_weight, None, grad_bias, None, None, None
+
+    @staticmethod
+    def _compute_loss(
+        input_chunk,
+        weight,
+        target_chunk,
+        bias=None,
+        preference_loss_fn=None,
+        full_target=None,
+        ignore_index=-100,
+        beta=0.1,
+        compute_nll_loss=True,
+        **loss_kwargs,
+    ):
+        """
+        Compute the total loss for a chunk of input and target, while using an alignment/preference loss function.
+        Args:
+            preference_loss_fn (callable): Loss function to compute the loss on a chunk of input/target.
+            input_chunk (torch.Tensor): Chunk of input tensor. Shape: (2 * chunk_size, sequence_length, hidden_size).
+            weight (torch.Tensor): Weight tensor. Shape: (vocab_size, hidden_size).
+            target_chunk (torch.Tensor): Chunk of target tensor. Shape: (2 * chunk_size, sequence_length).
+            bias (torch.Tensor, optional): Bias tensor. Shape: (vocab_size,).
+            full_target (torch.Tensor): Full target tensor. Shape: (batch_size, sequence_length).
+            ignore_index (int): Index to ignore for loss computation.
+            beta (float): Weight for the odds ratio loss.
+            loss_kwargs (dict): Additional arguments for the loss function.
+        """
+        len_chosen_chunk = target_chunk.shape[0] // 2
+
+        logits_chunk = input_chunk @ weight.t()  # chunk_size x V
+        if bias is not None:
+            logits_chunk = logits_chunk + bias
+        log_probs_chunk = F.log_softmax(logits_chunk.float(), dim=-1)
+
+        chosen_nll_loss = 0.0
+        if compute_nll_loss:
+            chosen_nll_loss = F.nll_loss(
+                log_probs_chunk[:len_chosen_chunk].view(-1, log_probs_chunk.shape[-1]),
+                target_chunk[:len_chosen_chunk].view(-1),
+                reduction="sum",
+                ignore_index=ignore_index,
+            )
+            chosen_nll_loss = (
+                chosen_nll_loss
+                / (full_target[: full_target.shape[0] // 2] != ignore_index).sum()
+            )
+
+        loss_mask = target_chunk != ignore_index
+        label_chunk = torch.where(loss_mask, target_chunk, 0)
+
+        per_token_logps = log_probs_chunk.gather(-1, label_chunk.unsqueeze(-1)).squeeze(
+            -1
+        )
+        average_log_prob = (per_token_logps * loss_mask).sum(-1) / loss_mask.sum(-1)
+
+        chosen_logps = average_log_prob[:len_chosen_chunk]
+        rejected_logps = average_log_prob[len_chosen_chunk:]
+
+        alignment_loss = preference_loss_fn(
+            chosen_logps, rejected_logps, beta=beta, **loss_kwargs
+        )
+        alignment_loss = alignment_loss / (full_target.shape[0] // 2)
+
+        loss = chosen_nll_loss - alignment_loss
+        return loss, (alignment_loss, chosen_logps, rejected_logps)

liger_kernel_nightly-0.4.1.dev20241114155849/src/liger_kernel/chunked_loss/orpo_loss.py

@@ -0,0 +1,63 @@
+import torch
+import torch.nn.functional as F
+
+from liger_kernel.chunked_loss.fused_linear_preference import (
+    LigerFusedLinearPreferenceBase,
+)
+
+
+class LigerFusedLinearORPOFunction(LigerFusedLinearPreferenceBase):
+
+    @staticmethod
+    def preference_loss_fn(chosen_logps, rejected_logps, beta=0.1):
+        """
+        Compute odds-ratio loss.
+        Args:
+            chosen_logps (torch.Tensor): Avg log probabilities of chosen tokens. Shape: (batch_size,).
+            rejected_logps (torch.Tensor): Avg log probabilities of rejected tokens. Shape: (batch_size,).
+            beta (float): Weight for the odds ratio loss.
+        """
+        log_odds = (chosen_logps - rejected_logps) - (
+            torch.log1p(-torch.exp(chosen_logps))
+            - torch.log1p(-torch.exp(rejected_logps))
+        )
+        ratio = F.logsigmoid(log_odds)
+        return beta * ratio.sum()
+
+    @staticmethod
+    def forward(
+        ctx,
+        _input,
+        weight,
+        target,
+        bias=None,
+        ignore_index=-100,
+        beta=0.1,
+        compute_nll_loss=True,
+        compiled=True,
+    ):
+        """
+        Fused linear layer with ORPO (Odds-Ratio Preference Optimization) loss.
+        Handles both the forward and backward pass of the final linear layer with ORPO loss.
+        Inspired from LigerFusedLinearCrossEntropyFunction (https://arxiv.org/abs/2410.10989) which fuses final linear layer and CE loss.
+        """
+
+        return LigerFusedLinearPreferenceBase.forward(
+            ctx=ctx,
+            _input=_input,
+            weight=weight,
+            target=target,
+            bias=bias,
+            loss_fn=LigerFusedLinearORPOFunction.preference_loss_fn,
+            compute_nll_loss=compute_nll_loss,
+            ignore_index=ignore_index,
+            beta=beta,
+            compiled=compiled,
+        )
+
+    @staticmethod
+    def backward(ctx, grad_output):
+        # Get gradients for _input, weight, bias, and target from the base class
+        grads = LigerFusedLinearPreferenceBase.backward(ctx, grad_output)[:4]
+        # Return these gradients, followed by None for the remaining inputs
+        return *grads, None, None, None, None

{liger_kernel_nightly-0.4.1.dev20241113011623 → liger_kernel_nightly-0.4.1.dev20241114155849/src/liger_kernel_nightly.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: liger_kernel_nightly
-Version: 0.4.1.dev20241113011623
+Version: 0.4.1.dev20241114155849
 Summary: Efficient Triton kernels for LLM Training
 License: BSD 2-CLAUSE LICENSE
         Copyright 2024 LinkedIn Corporation 

{liger_kernel_nightly-0.4.1.dev20241113011623 → liger_kernel_nightly-0.4.1.dev20241114155849}/src/liger_kernel_nightly.egg-info/SOURCES.txt

@@ -3,6 +3,9 @@ NOTICE
 README.md
 pyproject.toml
 src/liger_kernel/env_report.py
+src/liger_kernel/chunked_loss/__init__.py
+src/liger_kernel/chunked_loss/fused_linear_preference.py
+src/liger_kernel/chunked_loss/orpo_loss.py
 src/liger_kernel/ops/__init__.py
 src/liger_kernel/ops/cross_entropy.py
 src/liger_kernel/ops/fused_linear_cross_entropy.py